gather: deterministic context-pack tool (aft_gather)

[iceteaSA]

gather: deterministic context-pack tool (aft_gather)

What

One new tool — aft_gather — assembles a bounded "context pack" (ranked, deduped, budgeted verbatim code evidence) in a single call. It replaces the multi-turn search → outline → zoom → callgraph read chain an agent otherwise runs to build context around a question or a symbol.

Two modes (mutually exclusive):

• question: "how does X work?" — seeds from handle_semantic_search (same pipeline as aft_search, all lanes/fallbacks)
• symbol + filePath — seeds from the callgraph (impact depth-1 callers + call_tree depth-1 callees)

Seeds expand one hop through the callgraph, dedupe by canonicalized (file, symbol) with seeds winning, and render via render_symbol_within_budget until a hard line budget (default 400, cap 800) is spent. Everything past the cut appears as one-line stubs under ## Beyond budget (zoom to expand) — nothing is silently dropped.

Why

Agents burn serial turns assembling context: search, then outline the hits, then zoom the symbols, then chase callers. Each turn round-trips through the model. A pack returns evidence (verbatim bodies with file:line headers), not conclusions — the agent reasons over it directly, ready to attach to a subagent dispatch.

Measured on a real config repo (3 questions, one call each vs. the manual chain):

• "how does the lane-verdict cache flow end-to-end" → complete 4-file chain (cache shapes, verdict logic, write/read, plugin deny hook) in one pack, used=283/400. Manual baseline: 5-6 tool calls.
• "how does score-tap decide when to write an executor row" → the full decision chain (event filter → dedup → context sources → verdict parse → DB insert) in one pack at budget=200.
• An unrehearsed cross-file question → correct core symbols plus their docstrings, first try.

Independently reproduced on the aft codebase itself: question: "how does bash output compression dispatch pick a compressor" → seeds=15, used=226/400, one pack assembling the full dispatch chain (compress → gate → compress_with_registry_exit_code 20-compressor array → Compressor trait → install path → subc mirror) that otherwise takes a 4–5-call search→zoom chain.

Honest degradation

The pack never lies about its own quality:

• While the semantic index is building, long NL queries degrade to lexical-only file-level hits. The pack renders them as visible file:line (no containing symbol) stubs and flags the header with degraded=semantic-index-building (partial results — retry when index ready) — detected via the response's semantic_status field, cleared as soon as one real seed resolves. No blocking or retry inside the tool.
• Grep-fallback hits ({file, line_text, line}, no symbol name) resolve to their containing symbol by line containment — definitions, call sites, and comment hits all upgrade to the enclosing symbol. Hits with no containing symbol stay visible as stubs.
• Unresolved external/stdlib callees collapse to one summary line ((N unresolved external calls omitted)) instead of drowning the stub list; unresolved seeds and callers are never suppressed.

Implementation

• crates/aft/src/commands/gather.rs — Rust-side composition: calls handle_semantic_search / impact_result / call_tree_result / render_symbol_within_budget directly (shared &AppContext, no bridge round-trips, no parallel reimplementation of search).
• Wiring: main.rs dispatch arm, subc_translate.rs mapping, TS factory packages/opencode-plugin/src/tools/gather.ts + registration (same tier as aft_callgraph — depends on the callgraph store).
• No new dependencies, no config surface beyond the tool args, no LLM calls, no caching.
• Follows the tri-state honest-reporting convention (protocol.rs Response doc-comment): success:false+code for un-performable calls (e.g. invalid_request on a bad mode combo), success:true with a visible degraded/stub pack for partial results — never a bare empty success.

Tests

22 unit tests in gather.rs, including red-checked regressions (each confirmed to fail against pre-fix code): mid-codepoint truncation panic, duplicate-symbol line-anchored resolution, abs/rel path dedupe, containing-symbol resolution via a real TreeSitterProvider, callee-only stub suppression driven through the production build_pack path, and degradation-flag presence/absence/mixed cases.

Limitations (deliberate scope)

• 1-hop expansion only — multi-hop was deliberately excluded: the budget math and ranking get harder and the packs get noisier.
• Callgraph-dependent by design: neighbor expansion quality follows the callgraph store's freshness, same as aft_callgraph.
• Budget counted in lines, not tokens — matches aft's existing budget idiom across zoom/outline.

---

^{Need help on this PR? Tag /codesmith with what you need. Autofix is disabled.}

---

Summary by cubic

Adds aft_gather, a single-call, deterministic context‑pack builder that returns ranked, deduped, verbatim code within a fixed line budget, replacing the multi-step search→outline→zoom→callgraph chain. Exposed as gather in Rust and aft_gather in @opencode on the "all" surface.

• New Features

  • Two modes (XOR): question (semantic seeds) or symbol+filePath (impact callers + call‑tree callees), 1‑hop expansion, (file, symbol) dedupe with seeds winning.
  • Hard line budget (default 400, max 800) with per‑symbol balancing; overflow appears under “Beyond budget”; unresolved external callees collapse to one summary; unresolved seeds/callers stay visible.
  • Grep‑fallback hits resolve to containing symbols; no‑symbol hits render as visible stubs so nothing is dropped.
  • Honest header flags: degraded=semantic-index-building and neighbors=skipped(callgraph-unavailable); exact used lines computed post‑render.
  • Robustness: repo‑relative path resolution (incl. translate filePath), abs/rel normalization, Unicode‑safe query truncation, best‑match selection for same‑name symbols by start_line. 22 tests.
  • Tool surface: added to @opencode (ALL_ONLY_TOOLS, inventory updated); docs updated to include gather flow.

• Migration

  • Call with { question } or { symbol, filePath }; optional budget 1–800 (default 400).
  • Neighbor expansion requires the callgraph store in symbol mode; in question mode, neighbors are skipped with a header notice if unavailable.

^{Written for commit 66e05e2. Summary will update on new commits.}

Greptile Summary

This PR introduces aft_gather, a one-call deterministic context-pack builder that replaces the multi-turn search → outline → zoom → callgraph chain. The implementation is solid, well-tested, and follows existing codebase conventions throughout.

• New gather.rs command (1460 lines, 22 tests): seeds from semantic search (question mode) or callgraph impact + call_tree (symbol mode), deduplicates by canonicalized (file, symbol), renders within a hard line budget via render_symbol_within_budget, and emits over-budget candidates as visible stubs under ## Beyond budget. Path normalization, Unicode-safe truncation, and the callgraph store's own path resolution all interoperate correctly.
• TS adapter gather.ts: mirrors the Rust-side mode-XOR validation, enforces min(1)/max(800) on budget, and gates the tool in ALL_ONLY_TOOLS — consistent with aft_callgraph.
• Two minor gaps in gather.rs: when the search returns a completely empty results array while the index is building, the "no results found" response omits the degradation flag; the third dedup loop in build_pack for seed_ordinal.is_none() neighbors is currently unreachable since all generated neighbor candidates carry an assigned ordinal.

Confidence Score: 5/5

Safe to merge. The new gather command is well-isolated, carries no new dependencies, and the two gaps found are narrow edge cases that don't affect the core rendering or ranking path.

The implementation correctly integrates with the callgraph store (which normalizes relative paths internally, so question-mode neighbor expansion works as expected), the Unicode truncation fix is covered by regression tests, and budget accounting and header construction are provably correct after prior-round fixes. The two remaining issues are observable only in unusual conditions and carry no data-loss or security risk.

Files Needing Attention: The fully-empty-results branch in handle_gather_question (gather.rs ~line 286) and the third dedup loop in build_pack (gather.rs ~line 479) are the two spots worth a second look before shipping.

Important Files Changed

Filename	Overview
crates/aft/src/commands/gather.rs	New 1460-line core implementation of aft_gather; well-structured with 22 tests, two minor logic gaps: a missing degradation flag in the fully-empty results path, and a dead neighbor-ordering loop that can never execute.
packages/opencode-plugin/src/tools/gather.ts	Clean TypeScript adapter for aft_gather; duplicates mode-validation logic from translate_gather correctly, enforces budget min(1)/max(800) on the TS side, and follows the same pattern as other all-only tools.
packages/opencode-plugin/src/tool-registration.ts	Correctly adds aft_gather to ALL_ONLY_TOOLS and unconditionally spreads gatherTools(ctx) before the surface filter removes it for non-all surfaces, consistent with aft_callgraph.
crates/aft/src/subc_translate.rs	translate_gather correctly validates the XOR mode constraint, distinguishes missing-filePath from wrong-mode errors, and resolves filePath against project_root before forwarding.
crates/aft/src/main.rs	Minimal wiring change: adds a single dispatch arm for 'gather' adjacent to 'impact'.
docs/v0.49-unified-tool-surface-inventory.json	Correctly adds aft_gather to REG-V049-OC-ALL set and registers a HOSTONLY exemption for the OpenCode-only tool.

_{Reviews (10): Last reviewed commit: "gather: deterministic context-pack tool ..." | Re-trigger Greptile}

[cubic-dev-ai]

4 issues found across 7 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[iceteaSA]

Follow-up on the two maintainability notes from the Greptile summary (they weren't separate review threads, so noting here) — both addressed in cf09b9b9:

• degraded expression flagged as always-false in the non-empty-seeds path — intentional: degraded only applies when zero symbol-level seeds resolve (a building index yields no seeds; a ready index with seeds is never degraded). Logic unchanged; added a comment on the why so it doesn't read as dead code.
• callee-suppression string-match coupling — extracted UNRESOLVED_MARKER ("(symbol not resolved)") and CALLEE_PROVENANCE_PREFIX ("callee-of-") consts, referenced by both the producer (render_symbol_section err arm / collect_callees_for_seed) and the consumer (build_pack suppression guard) so the match can't silently drift. Only test-site literals remain, intentionally — a test pointing at the const couldn't catch const drift.

Matched/rendered text is byte-identical; 23/23 gather tests green.